Jumping on the Train of Personalized Medicine: A Primer for Non-Geneticist Clinicians: Part 1. Fundamental Concepts in Molecular Genetics.

With the decrease in sequencing cost and the rise of companies providing sequencing services, it is likely that personalized whole-genome sequencing will eventually become an instrument of common medical practice. We write this series of three reviews to help non-geneticist clinicians get ready for the major breakthroughs that are likely to occur in the coming years in the fast-moving field of personalized medicine. This first paper focuses on the fundamental concepts of molecular genetics. We review how recombination occurs during meiosis, how de novo genetic variations including single nucleotide polymorphisms (SNPs), insertions and deletions are generated and how they are inherited from one generation to the next. We detail how genetic variants can impact protein expression and function, and summarize the main characteristics of the human genome. We also explain how the achievements of the Human Genome Project, the HapMap Project, and more recently, the 1000 Genomes Project, have boosted the identification of genetic variants contributing to common diseases in human populations. The second and third papers will focus on genetic epidemiology and clinical applications in personalized medicine.

INTRODUCTION

Most human diseases have a genetic component. Non-genetic clinicians are familiar with single-gene disorders for the simple reason that the medical training, including human genetic courses, mainly refers to Mendelian diseases. An example of single-gene disorder is Huntington disease which is caused by a single mutation in the HD gene and that follows the easily recognized pattern of autosomal dominant inheritance across generations [1]. Some clinicians are, however, less comfortable with the principles of genetic contributions to complex disorders, despite the fact that a majority of human diseases (e.g. diabetes, cardiovascular diseases, cancers and psychiatric disorders) fall into this category. Complex diseases are triggered by multiple genetic variants in multiple genes acting in combination with environmental factors, and they typically do not follow any Mendelian patterns of inheritance. This limited knowledge in the medical community is understandable as genetic determinants for complex diseases were uncovered in the last 15 years and new discoveries are ongoing. Two important breakthroughs have revolutionized the search for genetic variants contributing to complex diseases and have boosted the elucidation of complex traits in the last seven years. First, the commercialization of high throughput genotyping microarrays has led to the emergence of genome-wide association studies (GWAS) and to an unparalleled harvest of disease-associated loci [2, 3]. Since the first report of GWAS in 2005, more than 2000 loci have been conclusively associated with one or more complex traits [4, 5]. However, most genetic variants from GWAS can only be correlated with a disease and the underlying mechanism may not be known. Over the past three years, the advent of high-throughput next generation sequencing platforms has led to the availability of whole-exome sequencing experiments which specifically sequence the subset of the human genome that code proteins, and to tremendous progress in the elucidation of Mendelian and complex disorders [6, 7]. With the decrease in sequencing cost and growing patient willingness to participate [8], personalized whole-genome sequencing may eventually become an instrument of common medical practice [9, 10]. These new perspectives challenge the clinicians to jump into the fast-moving field of personalized medicine, an emerging practice that uses an individual's genetic profile to guide decision-making in regard to the prevention, diagnosis, and treatment of diseases [11]. Despite all the recent ‘buzz’ around personalized medicine, the potential benefits of genetics in clinical practice are regarded with a certain degree of skepticism by the majority of clinicians [12, 13]. Obvious reasons include ethical concerns about privacy and discrimination or negative consequences of genetic testing for the patients (worry and anxiety) [14]. A less acknowledged but important reason is that genetics is regarded as a hermetic scientific field. The fact that geneticists use a highly technical language with terms like GWAS, single nucleotide polymorphism (SNP), and haplotype, certainly does not help. Ignorance begets fear and clinicians lacking the scientific background in genetic epidemiology may be prone to mistrust or scorn the promises and potential applications of genetic discoveries in their fields. Taking that into account, now is the time for clinicians to become more familiar with the key concepts of genetic epidemiology in order to become active participants of the personalized medicine revolution. We intend to write this series of three reviews to help non-geneticist clinicians prepare for the major genetic breakthroughs that will to occur in the coming years, and to welcome genomic medicine into their spheres of practice with the hope of achieving better prevention and care of human genetic disorders. In this Part I of the series of two reviews, we will detail the basic concepts of molecular genetics in user-friendly language. In the next two review we will then discuss the study designs and statistical procedures classically used in genetic epidemiology (in Part II) (and the realistic promises and challenges in application of recent genetic discoveries in medicine (in Part III)).

DNA, RNA AND PROTEINS

It has been known since immemorial time that offspring inherit, to a certain extent, their appearance, characteristics, and personality from their parents. As early as 1920s, a chemical substance called DeoxyriboNucleic Acid (DNA) was identified to carry the genetic information and transmit these characteristics from one generation to another [15]. In 1953, James Watson and Francis Crick described the double helix model of DNA structure which is the fundamental discovery for the central dogma of molecular biology [16]. Nucleotides are the basic units of the complex DNA molecule. They contain three parts of a five-carbon sugar, a phosphate molecule and a nitrogen-containing base, which is either adenine (A), thymine (T), cytosine (C) or guanine (G). Nucleotides polymerize into long chains by phosphodiester bonds (Fig. ). Because phosphodiester bonds link the 3’ carbon atom of one sugar to the 5’ carbon atom of the next sugar, the 5’ end has a terminal sugar residue in which the 5’ carbon atom is free and the 3’ end has a terminal sugar residue in which the 3’ carbon atom is free. Cellular DNA forms a double stranded helix (Fig. and ). The two coiled long polynucleotide chains are in an antiparallel formation in which the sugar-phosphate backbones are on the outside of the double helix, and the nitrogenous bases are on the inside and perpendicular to the backbones. Therefore, one strand runs in the direction of 5’ to 3’, whereas the other runs from 3’ to 5’. The hydrogen bonds between pairs of bases join the two strands following a specific and base-pairing rule: A to T only and C to G only. Although most of a cell’s DNA in humans is contained in the nucleus, mitochondria have their own independent genome that bears a strong resemblance to bacterial genomes [17]. Every cell in the human body has a complete set of DNA called a genome with the exception of mature red blood cells (erythrocytes), which lack a nucleus and most organelles. A gene is a segment of DNA along the genome encompassing specific regulatory elements (5’-untranslated regions (5’-UTRs) and 3’-untranslated regions (3’-UTRs)), non-coding regions (introns) and coding-regions (exons) which give instructions to messenger ribonucleic acid (mRNA) in the form of three base-pair sets called codons that assemble amino acids to a functional protein (Fig. ). Therefore, a gene is considered to be the basic unit of heredity. During cell growth and division, DNA replication initiates when two DNA strands unwind at a specific origin and serve as their own templates and synthesize the second copy of each DNA strand with the assistance of DNA polymerase and other enzymes. DNA self-replication is conducted in an extremely accurate manner (less than 1 mismatched nucleotide in 107) [18, 19]. Once an error occurs, a repair system including DNA polymerases, exonuclease and other enzymes will proofread DNA sequence and excise the incorrect base pair, ensuring the stability and high fidelity of DNA within an individual and across generations [20, 21]. On the other hand, if the repair system fails, a mismatch will lead to a de novo mutation. The DNA composition of the different types of cells in human is basically identical. However, the extent to which a given gene is “converted into” a functional protein may vary greatly in different cell types or even in the same type of cells at different states. Generally speaking, DNA is the instruction book, RNA is a photocopy of a specific page of the book and this page tells the cell how to make the protein. The RNA step ensures that energy is not wasted because the entire book is contained in every cell in the body, but certain cells only need to read certain pages (e.g. a nerve cell and a muscle cell use different sets of genes). When a specific protein is required, a process of trans- cription, the first step of gene expression, is initiated in which DNA is copied into an intermediate molecule named ribonucleic acid (RNA). One of the DNA strand serves as a template (called template or antisense strand), and RNA synthesis is also oriented in a 5’ to 3’ direction (corresponding to the N-terminus to C-terminus of the sequence of a polypeptide). The RNA transcript is complementary to the template, just like during DNA replication, except that a nucleobase uracil (U) pairs with A; therefore it has the same sequence as the non-template strand of DNA (which is called sense strand) except that U replaces T. A proofreading mechanism is also involved in transcription, but it is not as accurate as that of DNA replication [19]. Corresponding introns (non-coding sequencing in the RNA transcript) in a newly synthesized RNA molecule are subsequently removed by RNA splicing and a final mature messenger RNA (mRNA) is produced, which contains only exons (sequences that directly code for amino acids). The mRNA needs to be exported into organelle called ribosomes in the cytoplasm where the proteins are assembled. The sequence of an mRNA molecule is the template used to synthesize the corresponding a protein. The process by which mRNA is converted into a linear sequence of amino acids is called translation, the second step of gene expression. Specific nucleotide triplets, called codons, on the mRNA determine the start, the stop or the addition of an amino acid, leading to the creation of a polypeptide chain. Then a transfer RNA (tRNA), carrying the anticodon sequence (complementary to the codon on the mRNA) and a corresponding amino acid, binds to the codon on the mRNA and delivers the new amino acid to extend the polypeptide being synthesized. The maximum number of combinations of three bases out of four is theoretically 43=64 (Table ). Except for one specific codon (AUG) that initiates the translation of mRNA into protein and 3 codons (UGA, UAG, UAA) that stop translation, 61 out of the 64 triplets encode 20 different amino acids. Most amino acids are represented by more than one codon (e.g., six codons of UUA, UUG, CUU, CUC, CUA, and CUG for leucine). A specific codon always encodes a specific amino acid except in the case of the mitochondrial genome, which has four codons used differently from the nuclear DNA. This determines two important characteristics of the genetic code: specificity and degeneracy (also termed as redundancy). The degeneracy makes the protein more tolerant to some point mutations in coding regions and accounts for synonymous coding mutations. This means that a substitution of one nucleotide by another nucleotide does not necessarily result in an amino acid change (synonymous mutation) but others do change the coding sequence (non-synonymous mutations). All kinds of biological functions need the participation of proteins. However, the physiological roles of a protein depend on its amino acid sequence, configuration, and modulations from other relevant factors such as regulator proteins, ligands/ receptors or substrates. Mutations outside of the coding regions of the gene of interest may rather influence its mRNA expression or stability.

CHROMOSOMES, MITOSIS AND MEIOSIS

Most normal human somatic cells are diploid, and in their nucleus there are 46 continuous DNA molecules and each of them is named a chromosome [22]. The 46 chromosomes make up two sets, and therefore two copies of each chromosome have the same length, same centromere and identical genes and are designated homologs. One homolog is maternally inherited and the other is paternally inherited. Each set has 23 single chromosomes-22 autosomes and an X or Y sex chromosome. A male has an X and Y chromosome pair and female has a pair of X chromosomes (Fig. ). The twenty-two autosomes have been ordered from chromosome 1 to 22 according to the length of DNA base pairs (from the longest to the shortest). The X chromosome is much larger than the Y chromosome. The DNA sequences of two homologous chromosomes are usually not completely identical. DNA in a chromosome is packed in many complex units called nucleosomes consisting of two copies of core histones H2A, H2B, H3 and H4 around which is wound by a fragment of DNA, like many beads (histones) on a string (DNA). A fifth histone H1 is located in the spacer region between any two nucleosomes. Histone H3 and H4 can be modified by post-translational regulation mechanisms such as methylation, acetylation, ubiquitination and phosphorylation [23, 24]. These proteins are involved in epigenetic mechanisms, which determine in part the stable gene expression pattern from cell to cell or from generation to generation in the absence of change to DNA sequences [25, 26]. Along each chromosome, a constriction point called the centromere divides the chromosome into two arms: the shorter arm or “p arm” and the longer arm or “q arm”. In addition to identifying genetic diseases based on the patterns of G-banding (stained by Giemsa’s solution at the metaphase) [27], chromosome arms are useful to describe the location of a specific gene mutations.

There are two types of cell divisions, mitosis and meiosis [28]. Mitosis occurs in the context of body growth, cell differentiation, self-renewal and regeneration in somatic cells. DNA replication and partitioning go along with mitosis, ensuring the maintenance of a diploid chromosome stock (2×23 chromosomes) in daughter cells. Meiosis is a specialized reductive cell division which occurs exclusively in germ cells and gives rise to sperm and egg cells. In a single diploid spermatocyte or oocyte, DNA duplication generates two identical sister chromatids, followed by two DNA segregations and cell divisions known as meiosis I and II. During meiosis I, the homologous chromosomes, which are paired together to form a bivalent may possibly exchange a fragment of DNA between maternal and paternal strands. This process of exchange of genetic material is called recombination (crossover) and is one of the key mechanisms by which genetic diversity between daughter cells is generated. Subsequently, a complete set of 2×23 chromosomes are pulled to either pole and separated to form two haploid cells, each with one of the homologs. Which homolog in a bivalent pair ends up to in which daughter cell is independent and this is called the independent assortment. Independent assortment is the second major mechanism of genetic diversity. Therefore in humans, the total number of possible combinations of chromosomes in one gamete is 223. Meiosis II is similar to mitosis except that final daughter cells have 23 chromosomes instead of 46. As a result, meiosis eventually produces four haploid gametes. All eggs have a 23,X chromosome constitution representing 22 autosomes plus a single X chromosome, and 50% of the sperms have a 23,X chromosome constitution and the other half are 23,Y (Fig. ). When a sperm fuses to an egg, a zygote is formed and the diploid chromosomal status is re-established. Taken together, the daughter cells from mitosis are genetically identical, whereas the daughter cells from meiosis are genetically different as a consequence of independent assortment and recombination. As discussed earlier, de novo DNA mutation may be caused by a failure in the repair system during DNA replication. In addition to this, an error in combination process may also generate structure abnormalities in DNA. The average number of crossovers per cell is about 55 in males and is approximately 50% more in females, which means crossovers are not rare events. Crossovers are essential in maintaining the genetic variability that is transferred from parent to offspring. The exception again is mitochondrial DNA, which is inherited as a single linked molecule through the female line. It does not undergo recombination. Just as in DNA replication, errors during recombination do occur at a very low frequency, giving rise to translocations, inversions, duplications, or deletions.

Abnormalities of chromosome structure are reported to contribute to a small portion of cases in psychiatric disorders. Balanced translocation is an exchange of chromosome segments between two non-homologous chromosomes. A balanced translocation between chromosome 1 and 11 disturbed DISC1 gene is associated with increased risk of schizophrenia [29, 30]. Chromosome inversion occurs when there are two breaks in one chromosome and the same segment is re-constituted with the orientation inverted. A pericentric inversion on chromosome 9 was found to be associated with schizophrenia [31].

CHARACTERISTICS OF THE HUMAN GENOME

The completion of the Human Genome Project in April 2003 and the 1000 Genomes Project in 2012 has revealed several important characteristics of the Human genome [32, 33]: 1) there are about 3 million base pairs in the human genome; 2) 99% of nucleotide bases are the same in all humans; 3) an estimated 30,000 genes exist in humans, with an average length of 3000 base pairs; 4) genes represent less than 2% of the human genome; 5) more than 50% of genomic DNA consist of non-repetitive DNA sequences and most of the genes display unique DNA sequences; 6) about 45% of genomic DNA consists of repetitive sequences which are thought to contribute to maintaining chromosome structure; 7) there are 38 million validated SNPs in which a single nucleotide differs at a particular position among 1,092 human genomes.

GENETIC VARIATIONS

The current entire database of human genomic variation was recently derived from a panel of whole-genome sequence data in 1,092 individuals from 14 populations in the context of the 1000 Genomes Project [33]. The next targeted milestone of the 1000 Genomes Project is sequencing the genome of 2,500 individuals from 27 populations across the world [34]. Although 99% of the genomic DNA sequences are identical, 1% still signifies 38 million genetic variants between unrelated individuals, indicating there is one allele variant in every 80 base pairs on average. During the assembling of consensus sequences, differences between (among) the nucleotide sequences of different individuals were noticed. SNPs represent more than 90% of all human variation [35]. If the frequency of a SNP is greater than 5%, it is considered a common variant or polymorphism. If the frequency is between 1-5%, it is a low-frequency SNP. If the frequency is less than 1% in population, it is defined as a mutation. In addition to SNPs, other genetic variants have been observed in the human genome, including microsatellites, minisatellites, and copy number variants (CNV). The 1000 Genomes Project also identified 1.4 million bi-allelic short insertions and deletions, and more than 14,000 large deletions [33]. Because these genetic variants were discovered during the sequence assembly, their locations are inherently known, providing a key resource in mapping genes that predispose to common diseases.

Genetic variants may occur in any region of the genome. A SNP that is located in the coding region without changing the corresponding amino acid is called synonymous, while coding SNPs that lead to changes of the amino acid, shifting of the reading frame or to an earlier stop code are called non-synonymous, frameshift or non-sense, respectively. SNPs found in a non-protein coding area in a gene may influence the protein expression by changing regulatory elements such as transcription factor, binding sites or configuration. In humans, there are usually only two alleles at a SNP location, but three alleles are sometimes reported, such as e2, e3, e4 alleles at the APOE gene locus [36]. The most common nomenclature of a SNP uses a unique reference SNP (rs) number. An example is the rs10994336 SNP in the ANK3 gene that has been associated with bipolar disorder [37]. SNP data are available from publicly accessible resources and are constantly updated, such as dbSNP polymorphism repository, Human Genome Variation Database, the International HapMap Project, SNP consortium or 1000 Genomes Project database.

Microsatellites or short tandem repeats (STR) refer to repeated sequences of less than 10 bp of DNA. When the repeat units have 10-100 nucleotides and the copy number may reach up to a few thousand, this repeat cluster is referred as a minisatellite or a variable number tandem repeat (VNTR) [22]. The number of alleles in microsatellites and minisatellites is usually 5 or more. Though both microsatellites and minisatellites are highly unstable, the majority of the variations have no detrimental clinical consequences. The mutation mechanisms in microsatellites and minisatellites are different. In minisatellites, mutations occur during homologous recombination at meiosis, but the rate is approximately 10 times greater than that of other DNA sequences. Microsatellites undergo slip-strand mispairing during replication and subsequently the genes in the repair systems are inactivated, leading to expansion of the repeats [22]. This mutation rate is also several of orders of magnitude higher than the mutation processes that lead to SNPs. Some of them result in increased risks of diseases. For example, whereas healthy individuals carry less than 36 repeats of CAG in the HD gene, the number of repeats increases to more than 40 in individuals who will develop Huntington disease [1].

Another type of polymorphism is called copy number variant (CNV). CNV refers to the duplication or reduction of a DNA segment (200 bp to 1.5Mb) and they may have only two or multiple alleles. CNVs (deletion or duplication) can have important functional consequences and have been convincingly associated with psychiatric disorders such as schizophrenia [38].

ALLELES AND GENOTYPES

The location of a DNA sequence or a gene on a chromosome is called a locus. If there is more than one type of nucleotide at a specific locus in a population, each nucleotide is called an allele. Most polymorphic sites have only two alleles, while a few have more than two alleles. Individuals are called homozygotes when the two alleles of homologous chromosomes at a specific locus are identical. When the two alleles are different, individuals are classified as heterozygotes. At bi-allelic SNPs, the allele with higher frequency in a given population is called the major, and the less common one is called minor allele. The three (or more) possible combinations of alleles at a specific locus (e.g. major allele / major allele, major allele/ minor allele, minor allele / minor allele) are called genotypes. Sometimes, a genotype refers to the overall genetic constitution of an individual.

For instance, the SNP rs1024582 in the CACNA1C gene is associated with bipolar disorder and schizophrenia [39]. There are two alleles A and G, A being the minor allele with frequency of 33.7% and G being the major allele. The three genotypes of an individual at this locus can be AA, AG or GG. The minor allele A increases the risk of bipolar disorder and schizophrenia [39].

HAPLOTYPES AND LINKAGE DISEQUILIBRIUM

Alleles of different loci are sometimes not independently transmitted from one generation to another. They may be physically linked on the chromosome and the crossovers across generations do not break them apart. Such a cluster of alleles is called a haplotype (Fig. ). The US National Institute of Health initiated the International HapMap Project in 2002 to develop a human haplotype map [40]. In phase I more than 1 million common SNPs were genotyped in 2005 in 270 individuals from four geographically distinct populations, Japanese, Han Chinese, Yoruba of Nigeria and Americans of North Western European ancestry [41]. These data were used to explore the patterns of association among SNPs in the genome, and how these patterns vary across populations. In Phase II HapMap, over 3.1 million SNPs were genotyped to create a second generation human haplotype map [42].

Linkage disequilibrium (LD) measures the non-random association of alleles at two or more loci that may or may not be on the same chromosome. For instance, we may consider two loci with alleles A1/A2 and B1/B2, A1 and B1 alleles being on one chromosome and A2, B2 alleles being on the other homologous chromosome. The frequency of A1 is 60% and B1 is 30% in a population. If the recombination of the two loci is independent, the expected frequency of the four possible haplotypes A1B1, A1B2, A2B1 and A2B2 would be 18%, 42%, 12% and 21%, respectively. If the distribution of these four haplotypes is consistent with the theoretical frequency, the alleles are in linkage equilibrium. If the distribution significantly departs from the theoretical frequency, the alleles are in linkage disequilibrium (LD), indicating the two loci are not independent. If one of the alleles is a disease causing allele, the haplotype including this allele is considered as a disease-containing haplotype. In most circumstances, a genetic variant that is found to be associated with a disease is not the functional diseasing causing allele; rather it is a proxy SNP. This indicates that this proxy SNP is in the same LD block with the potential causal SNP which is not genotyped in the array. LD may change over time and the patterns of LD may vary depending on the population. The sizes of LD blocks, which reflect the frequency of recombination, have been reported to be smaller in African than in Asian and European populations (Fig. ) [43]. Thus, knowing the LD pattern in a specific ethnic group (e.g. from the HapMap Project) is useful to refine the association signal and ultimately lead to the discovery of the causal variant [44, 45]. Many other factors may influence LD patterns, including random genetic drift, population growth, admixture, inbreeding, natural selection, and de novo mutation [46].

Using the example of haplotype given above, one statistical test to measure LD is D’=D/Dmax=(PA1B1-PA1PB1)/ Dmax, where P represents the possibility of a specific haplotype and Dmax is the maximum difference between PA1B1 and PA1PB1. D’ ranges from -1 to 1. One or -1 denotes there is no recombination between two loci A and B, and 0 indicates that A and B are in linkage equilibrium. If the allele frequencies of A1 and B1 are similar, a high D’ value indicates A is a good surrogate for B. However, if the sample size is small or one allele is rare, D’ will be inflated. There is a second measurement of LD, using the squared coefficient of determination r2 (ranging from 0 to 1). r2 takes into account the sample size and allele frequency. Therefore, D’ is extensively used by population geneticist to assess recombination patterns such as defining haplotype patterns, whereas r2 is a more appropriate measure of linkage disequilibrium in association studies [47]. For example, two SNPs can display a D’ value of 0.85 and a r2 value of 0.18. In an association study, these two SNPs cannot be tagged or substituted for each other because of low r2. Pairwise measurement of LD for neighboring SNPs are used to group more than 2 loci into a haplotype termed LD block if the values of D’ between any two SNPs within the group are above a certain threshold (e.g. D’ > 0.8). This knowledge is essential to guide the design of whole-genome SNP genotyping arrays because carefully selecting a single or a few SNPs representing a haplotype block due to their strong LD can be used to identify an important haplotype, rather than genotyping all the SNPs in this haplotype (Fig. ). The common SNPs in commercial genotyping arrays captures untyped common variation with an average maximum r2 (a correlation coefficient between genotyped and untyped SNPs) from 0.9 to 0.96 depending on the population. Therefore, the advances from Phase II HapMap, in combination with increased density of high-throughput technology and capability of imputation of untyped SNPs, greatly improved the power of association studies. HapMap 3 was completed in 2009 and it genotyped 1.6 million common and rare variants including CNVs in 1,184 reference individuals from 11 global populations [48]. The integrated map of genetic variation from the complete HapMap data and the 1000 Genomes Project [33] enables analysis of common and rare variants and CNVs in populations of different ethnic background. For instance, Sung and colleagues recently derived the genotypic distribution of 6.7 million SNPs from the information of 324,607 SNPs genotyped in their sample, using the reference panel from the 1000 Genomes Project [49].

CONCLUSIONS

Having a stronger background in molecular genetics, we are ready to discuss the subtle concepts of genetic epidemiology including study design implementation, gene identification strategies, genetic marker selection, genotyping and sequencing technologies, data analyses, data interpretation and their potential applications in the context of personalized medicine. The two next article in this series will review this topic.

Table 1.

The genetic code.

Sixty-four different combinations of triplet codons are derived from 4 unique bases. Except ATG for the start codon and TAG, GTA, TAA for stop codons, each codon codes for one of the 20 amino acids.